7/31/2019 MPRA Paper 33024

1/40

Munich Personal RePEc Archive

Threatening to increase productivity

Bridgman, Benjamin; Gomes, Victor and Teixeira, Arilton

2010

Online at http://mpra.ub.uni-muenchen.de/33024/

MPRA Paper No. 33024, posted 27. August 2011 / 18:08

http://mpra.ub.uni-muenchen.de/33024/http://mpra.ub.uni-muenchen.de/33024/http://mpra.ub.uni-muenchen.de/

7/31/2019 MPRA Paper 33024

2/40

Threatening to Increase Productivity

Benjamin Bridgman Victor Gomes Arilton Teixeira

October 7, 2009

Abstract

The wave of privatization in the 1980s and 1990s increased productivityof many previously state owned enterprises (SOEs). However, governments

often do not have sufficient support to privatize SOEs. We provide evidence

that threatening privatization and market competition (entry of new firms)

can increase the productivity of SOEs, even though privatization and entry of

new firms does not occur. We study productivity at Brazils state-owned oil

company Petrobras. After it lost its legal monopoly Petrobrass total factor

productivity increased sharply. These large gains occurred despite the fact that

Petrobras faced no immediate de facto competition. The threat of competition

and privatization was sufficient to generate large productivity gains. Thesefindings suggest that changing the competitive environment can be a powerful

force for improving productivity at state-owned firms.

We thank Edward Prescott, Berthold Herrendorf, James Schmitz Jr., Ryan Greenaway-McGrevy, Marc Muendler and Bruno Funchal for comments and Patrcia Moura da Silva for researchassistance. Sergio Barros da Cunha, Demetrius Casteloes, and Marcelo Duque of Petrobras pro-vided valuable technical background. Philip Beckett and Mike Earp of the U.K. Department ofTrade and Industry kindly provided data. The views expressed in this paper are solely those of theauthors and not necessarily those of the U.S. Bureau of Economic Analysis or the U.S. Departmentof Commerce.

Bureau of Economic Analysis, Benjamin.Bridgman@bea.gov.Universidade de Brasilia, victorgomes@unb.br.Capixaba Research Foundation, arilton@fucape.br.

1

7/31/2019 MPRA Paper 33024

3/40

1 Introduction

There is a great deal of evidence that private firms perform better than public ones.

The wave of privatization of the 1980s and 1990s led to higher productivity at a num-ber of previously state-owned enterprises (SOEs). (See Megginson and Netter [28]

for a survey.)

However, governments are often politically constrained from privatizing. While

some governments would like to privatize these firms, they cannot. Privatization

often does not have public support and those who benefit from state ownership may

able to block it. Some high profile privatizations, such as those in Russia, were done

in a way that many people perceived as only benefiting politically connected elites.

People in countries with a colonial history are often leery of privatization since it

risks placing public assets in the hands of foreigners, thought of by many as neo-colonialism. Clear majorities of the population in many Latin American countries

do not think privatization is beneficial and its popularity has been falling over time

(McKenzie and Mookherjee [27]).

Are SOEs doomed to poor performance in cases where privatization is not politi-

cally feasible? In this paper, we show that governments can improve the performance

of SOEs even without privatization. We study a case where the threat of market

competition and privatization increased productivity and improved the performance

of a SOE, even though privatization and entry of new firms did not occur during the

period we examine.

In 1995, Brazil ended the legal monopoly rights of its state-owned oil company,

Petrobras, over production, refining, import and export of oil. Petrobras was not

privatized or broken up and the government claims that there are no plans to do

so. In fact, to get the Brazilian Congress to end monopoly rights of Petrobras

the government formally agree that privatization from then on could only be made

with the approval of Congress. We evaluate the productivity performance of the

companys oil extraction division before and after the loss of monopoly and the

change of the environment.

The main finding is the reform led to a significant increase in Petrobrass produc-

tivity growth. Between 1976 and 1994 labor productivity grew at an annual average

2

7/31/2019 MPRA Paper 33024

4/40

of 4.7 percent whereas between 1995 and 2001 it grew at an annual average of 14.6

percent. Not only did growth triple, the sources of the growth changed. Labor pro-

ductivity growth prior to 1995 was due to capital accumulation and materials use

with little TFP growth. Beginning in 1995 TFP grew rapidly, accounting for almostall the increase in the labor productivity growth rate.

To accomplish these results, Petrobras slashed its use of inputs while maintaining

output growth. It also began to shift its portfolio of oil wells to more productive

regions and changed its corporate structure. The speed with which the changes

were implemented indicate that they were feasible prior to reform (physically, if not

politically).

What makes this case particularly compelling is that labor productivity growth

immediately doubled despite the fact that the reform resulted in no immediate

changes in Petrobrass market or ownership. In the period we examine, no other

firms entered the market and no competing imports were allowed. The firm was not

been privatized and has not been in the 15 years since the reform. Petrobras main-

tains a dominant position, still extracting 97 percent of Brazils oil. Its advantages as

a large incumbent made it difficult for competitors to enter the market. The changes

brought about by the reform were sufficient to generate significant productivity gains.

The intuition for these changes in the performance of Petrobras is the following.

Petrobras was a monopolist and competing imports were prohibited (only Petrobras

could import oil). Shielded from competitors, Petrobras was not forced to minimize

costs. Many of the non-economic goals that it pursued, such as increasing employ-

ment, directly reduced productivity. Petrobrass production cost was higher even

when it used the best technology available.

After the reform, Petrobras knew that it would lose its monopoly over the Brazil-

ian market. Facing the prospect of competing against lower cost producers, Petro-

brass managers had an incentive to get rid of non-economic goals, reducing costs

even before the changes were implemented1. Had Petrobras continued to pursue

non-economic goals, it faced losing significant market share. Petrobras needed to

1

There is a literature suggesting that competition over a field (that is, more than one companyexploring the same field at the same time) can generate inefficient allocation (Libecap and Wig-gins [26]). As shown below, this did not happen in the Brazilian oil industry. By competition wemean here more than one company exploring different oil fields in Brazil at the same time.

3

7/31/2019 MPRA Paper 33024

5/40

reduce its costs to avoid losing its dominant position in the market.

The prospect of foreign firms in the market also constrained the government.

Knowing that lower cost firms would be entering, it could not ask Petrobras to

pursue non-economic goals without inducing losses that the government would haveto cover. It also helped commit future governments to maintain the open markets

by making a return to a monopoly more costly. Reclosing the market would require

expropriating Petrobrass foreign competitors, risking international retaliation, or

buying them out.

The reform also increased the threat of privatization. The constitutional change

lowered the institutional barriers to privatization. The legal requirement was reduced

from a Congressional supermajority to a simple majority. It was also a clear sign that

the government had changed his objectives. Since its creation, Petrobrass monopoly

rights had only been expanded. (We will discuss the history of Petrobras below).

The reform was the first time that Petrobrass legal monopoly had been raised in

Congress. Therefore, the constitutional change pursued by the government and ac-

cepted by Congress was a clear sign that the governments attitude had changed.

In addition, new producers in the Brazilian market would provide a performance

benchmark. If other companies operating in Brazils oil market were successful while

Petrobras was not, it would provide the government with a rationale for privatizing

Petrobras.

These results have some policy implications. When privatization is not politically

viable, increasing competition in the markets of state-owned firms can provide an

avenue for improving performance. Also, indicators of market outcomes can be poor

proxies of market competitiveness. The prospect of competition resulted in major

changes in Petrobrass management strategy and productivity. However, commonly

used market outcome proxies such as concentration indices changed very little with

the reform. As a proxy for competition, they would have missed a important shift

in the competitive environment.

Our findings are in line with that of Goolsbee and Syverson [19] and Bartel and

Harrison [3]. The first shows that the threat of entry in the U.S. airline industryled to price cuts by incumbents before the entry occurred. The second provides evi-

dence from Indonesia that the competitive environment is an important determinant

4

7/31/2019 MPRA Paper 33024

6/40

of productivity, regardless of ownership. Our findings are also consistent with con-

testable markets theory, which argues that a monopolists behavior is affected by the

threat of new entrants, not just actual entrants (Tirole [37]). Similar to Bresnahan

and Reiss [5], adding an entrant to a market with few incumbents has a large effect.Though only one company entered the oil extraction market in the first ten years of

the reform, the productivity growth rate more than doubled.

Our paper is part of a growing literature examining the productivity effects of

discrete changes in competitive pressure. Galdon-Sanchez and Schmitz [15] and

Schmitz [21] find that productivity increased sharply in response to increased com-

petitive pressure in the 1980s. Schmitz and Teixeira [22] find that private firms

productivity increased when public firms in the Brazilian iron ore industry were pri-

vatized. Cole, et al. [11] examine a number of these studies for Latin America. The

closest work to ours is Garcia, Knight and Tiltons [16, 17] study of Chilean copper

mining. Industry labor productivity increased after other firms were allowed to com-

pete with the state-owned monopoly Codelco. In contrast to our findings, most of

the increase came from high productivity entrants rather than improved productivity

at Codelco.

It is closely related to the barriers to riches literature. Parente and Prescott [31]

build a model where a monopolist sets price above marginal cost and does not use

the most productive technology. Our empirical results match their theoretical pre-

dictions. After the end of monopoly rights of Petrobras, output and technological

progress (as measured by TFP growth) increased sharply.

The causes of the post-reform productivity gains are similar to those found in Ol-

ley and Pakes [29]. They study the effects of deregulation and entry on productivity

at the plant level in the U.S. telecommunication industry. They find that productiv-

ity increased mainly due to reallocation of inputs (capital) toward more productive

plants. Productivity growth increased in a similar way at Petrobras, as it shifted

production to more productive wells after the loss of its monopoly.

Laitner [24] separates the costs of imperfect competition in two parts: a static

and dynamic effect. The static effect is the well-known deadweight loss caused bysetting price above marginal cost (Harberger Triangles). The dynamic effect is due

to underinvestment, reducing the capital stock and production. While Petrobrass

5

7/31/2019 MPRA Paper 33024

7/40

production grew rapidly after the reform, to the point that Brazil is a zero net im-

porter of oil, investment did not. Production was increased by using capital and other

inputs more efficiently. In fact, it was during the monopoly period that Petrobras

invested heavily, due in part to a national policy designed to encourage investment(Bugarin et al [9]).

Our work is largely consistent with the literature on the effects of competition.

However, in contrast to much of this work, we study a public company. We find that

public companies react to competition in the same ways as private ones. However,

the nature of the inefficiency of monopoly may be different, since public firms may

pursue political goals that privately held companies do not.

2 Petrobras and the Brazilian Oil IndustryPetrobras is an integrated state-owned oil company that extracts, imports and ex-

ports, and refines crude oil and distributes gasoline2. It is a major player in the

world oil industry and was ranked 125th in the 2005 Global Fortune 500. It is also

very important in Brazil. Its sales are 6 percent of Brazils GDP.

While its sheer size makes it economically important in Brazil, Petrobras is also

politically important. It is one of the Crown Jewels, a set of politically sensitive

state-owned companies that figure prominently in the nationalist movement. Taxes

on oil extraction are an important source of revenue for Federal, state, and citygovernments. There are also indirect benefits. It employs skilled, high-wage workers

and is a source of local supply contracts.

Petrobras was created in 1954 and given a monopoly over oil exploration, ex-

traction and refining. (Existing private refining companies were allowed to continue

operating but could not expand.) This policy was part of a larger import substitution

policy to develop the industrial sector (Kingstone [23]).

From its creation until the 1990s, Petrobrass control of the Brazilian oil market

steadily increased.3 The monopoly power of Petrobras was extended to the import

2The government holds a majority of the voting stock. Non-voting shares and a minority ofvoting shares have been held by the public since the creation of Petrobras.

3See Campos [10] and Barreto [2].

6

7/31/2019 MPRA Paper 33024

8/40

and export of oil in the 1963. Between the creation of Petrobras and 1970s, domestic

prices were essentially equal to the international price plus a Federal tax. In an

attempt to shield the domestic economy from the oil shocks, domestic prices became

disconnected from international prices in 1977 and based on a measure of domesticproduction cost. In 1988, Petrobrass monopoly rights were guaranteed in the new

Brazilian constitution.

The policy to remove the monopoly originated with President Fernando Hen-

rique Cardoso. The first time that he publicly mentioned his intention of ending

Petrobrass monopoly rights was in December of 1993 while he was still the Finance

Minister (Prado [33, page 178]). He was elected with a majority in the first round

of presidential voting in October 1994 and took office in January 1995. The election

also provided him with strong legislative support, which his reformist predecessors

did not have. Having strong legislative support was important for the reform since

changing the constitution requires a supermajority vote in the Congress. In February

1995, he sent the amendment to the Congress eliminating Petrobrass monopoly.

Congressional resistance to the reform was strong. Members from both the left

and right opposed it on ideological grounds; the left was opposed to private in-

volvement and the right opposed weakening a nationalist symbol. The reform also

threatened patronage opportunities. The reform likely would have failed without a

strike by Petrobrass employees that generated a shortage of natural gas, which is

used widely for cooking. Public opinion turned sharply against the strikers, which

coupled with his personal popularity as the architect of currency stabilizing Real

Plan, allowed Cardoso to push the reform through the Congress. Cardosos congres-

sional allies extracted a concession in the form of a clause prohibiting privatization

in the law regulating the oil sector4. (This Petroleum Law was approved in 1997, as

seen below). To end the monopoly, Cardoso had to give up the idea of privatizing

Petrobras (Kingstone [23]).

The amendment was approved in November 1995. The Federal government re-

tained the ownership of hydrocarbon reserves but opened the sector to private firms,

ending 40 years of monopoly. Below we rgue that the reform began in 1995 sinceonly after the election of president and the congress in October of 1994 Petrobrass

4The law requires that at least 51 percent of the voting shares be held by the Federal government.

7

7/31/2019 MPRA Paper 33024

9/40

managers would have seen Cardosos likely ability to pass reform upon his election

in that year.

Market opening was phased in. The amendment called for new legislation reg-

ulating the oil market. This law, called the Petroleum Law, was approved by theCongress in August 1997. It created the National Petroleum Agency (ANP hence-

forth), which took over regulation of the oil market from Petrobras. The Petroleum

Law set out the process of liberalization for all sectors related to production of oil

products. In this paper, we focus on the extraction sector. (For a description of the

deregulation of other sectors, see Serour [36]).

The Petroleum Law allowed Petrobras to keep its rights of extraction of oil in

areas where it could prove that it had done investment. In August 1998, Petrobras

signed 397 contracts with ANP, 282 of which covered areas already under develop-

ment. These contracts, called Round Zero, gave the rights for 450 thousand square

kilometers to Petrobras without payment. After that, new areas for exploration were

offered in auctions where any company from any country could participate. By 2003,

41 companies, besides Petrobras, had bought some areas. In that year, Shell was the

first private company to produce oil in Brazil, though Petrobrass share of Brazilian

production is still almost 100 percent.

In 2001, exports and imports of oil and its derivatives were opened to companies

besides Petrobras (through its subsidiary Transpetro).

In what follows, we date the beginning of the reform in the oil industry as 1995.

Even though the change in Petrobrass monopoly status occurred at the end of 1995,

as noted above, discussion of the policy change began earlier. When dating the

beginning of a reform, the date of the legal change may not be the most relevant date.

Managers in the reformed industry may have anticipated the reform and introduced

changes prior to the reform becoming official5. For example, in a study of the impact

of privatization in the Brazilian iron ore sector, Schmitz and Teixeira [22] argue that

the reform began in 1990, when a new President was elected promising to privatize

5We would date the beginning of the reform as the election of Cardoso and his congressionalallies in October 1994 if we had monthly data. We designate 1994 as a non-reform year since we onlyhave annual data. The election occurred late in the year and any changes Petrobrass managementmade would take time to take effect. We test the robustness of our results to including 1994 as areform year below.

8

7/31/2019 MPRA Paper 33024

10/40

and deregulate the economy, rather than the official announcement in 1995.

3 Production Process in the Oil IndustryThis paper examines oil and gas extraction productivity. This section discusses

the techniques employed to extract oil and gas from the soil and preparing it for

refining, with particular attention to the attributes of wells that affect extraction

productivity.6

Once a field with oil and gas is discovered, the first step is to install the wells:

holes where pipes are inserted to extract oil and gas from the rock. Recent advances

in satellite seismic imaging have improved exploration and the siting of wells. In

the 1990s, seismic imaging shifted from two to three dimensional geological images.

These techniques are available in the international market and any company in the

sector can use it.

Every well produces both oil and natural gas simultaneously.7 Extraction may

or may not use energy. Some wells have natural pressure that forces oil out without

additional inputs. Other wells do not have enough natural pressure, so it must be

built up artificially. This can be done by either pumping water into the reservoir or

compressing gas into the well.8 Adding pressure requires energy which comes from

the gas produced by the wells themselves, reducing net output. The consumption of

gas can reach 30 percent of the total produced by a well.The existence of natural pressure in the well is not related to any geographic

characteristics of the field or the well. Both deep sea and land based wells may or

may not have natural pressure. For example, wells in the Middle East have natural

pressure, making it a low cost producer.

Natural pressure is related to the age of the well since removing oil from a field

reduces its pressure. It is possible to reduce the amount of energy spent to produce oil

6This section draws heavily on discussions with Petrobras engineers Sergio Barros da Cunha,Demetrius Casteloes, and Marcelo Duque.

7Generally the separation of oil from gas is automatically done through the change of pressure.That is, inside the field there is high pressure. Once the oil and gas comes to the surface, thereduced pressure allows an automatic separation. See Worley and Laurence [38] and Rondeon [35].

8See Bennion et al [4] and Patton [32].

9

7/31/2019 MPRA Paper 33024

11/40

and gas by shifting production to new wells, even when the new wells are much bigger

than the old ones. As discussed below, this happened in the Brazil when Petrobras

abandoned older, smaller wells in the states of Alagoas, Sergipe, and Bahia and

concentrated its production in the state of Rio de Janeiro.The oil and gas extracted from wells comes mixed with water which must be

separated out. The oil and gas are separated from water using chemicals. The

amount of chemicals required may not increase in the same proportion to the liquid

extracted from the well. Chemical use depends on the temperature of the liquid once

it has reached the platform, with colder liquid using more chemicals.

Larger and newer wells may use fewer chemical inputs. First, the amount of water

coming out of a well together with oil depends on the age of the well. The older the

well, the more water it produces. Second, equipment specialized in the separation

of oil from water can be attached to a platform. It uses gas to produce energy

that is use to separate oil from water through a electrostatic treater9. This reduces

transportation costs, since less water is transported to the refinery. This technology

is only economical for wells that produce a high daily volume. This technology is

used in Rio, where wells are large, but is rare in the states where wells were closed.

We conclude that a shift to newer and larger wells used in production likely re-

duced materials usage per well. Newer wells tend to have higher pressure, so they

consume less energy than older wells, and produce less water, so they use fewer chem-

icals inputs. The move to larger and newer wells also allowed the use of separating

technology based in gas that reduces chemical use and reduce transportation cost.

4 Productivity at Petrobras

In this section, we analyze Petrobrass domestic oil extraction units labor productiv-

ity and TFPperformance in response to the loss of its legal monopoly. We argue that

the end of Petrobrass monopoly and the threat of new competitors had an impact on

its productivity performance. We compute Petrobrass TFP using a Cobb-Douglas

production function given by9See Cummings and Engelman [12] and Bromley, Gaffney and Jackson [6].

10

7/31/2019 MPRA Paper 33024

12/40

Yt = AtKtM

t N

(1)t (1)

where Kt is the aggregate capital stock, Mt is the amount of material, Nt is the

number of employees, is the capital share, is the labor share and At is TFP. We

first discuss the construction of the data series, followed by the choice of factor share

values.

The data are drawn from the ANP and the balance sheets of Petrobras. Output Yt

is the total physical quantity of oil produced. Natural gas produced, a joint product

of oil, is converted into oil equivalents as described in the Appendix. Labor input Nt

is the number of exploration and production employees for each year, adjusted for

contracting out as described in the Appendix. These data begin in 1976.The capital stock is calculated using perpetual inventory on capital expendi-

tures.10 Data extends back to the establishment of Petrobras, so we directly observe

the initial stock. (Further details are available in the Appendix). Capital expendi-

ture measured in the Brazilian currency was converted into U.S. Dollars using the

December monthly exchange rate and deflated by the U.S. oil and gas equipment

price index.

We did not have data for Petrobrass material use so we use wells in production

as a proxy. This assumes average material use per well is constant over the period

considered. The robustness of the results to this assumption and those used toconstruct the capital series are discussed below.

We now turn to the choice of factor share values. Petrobrass average labor share

in the period 1976 and 2001 is 0.2. Since we do not have data on Petrobrass materials

expenditures, we cannot calculate the capital and materials shares directly. Instead,

we use U.S. data. Labor share in Brazil is same as that of the United States, which

10Our measure of capital includes capital formation in two sectors - well exploration and devel-opment - but we do not include development as part of our output. Well development is fairlyconstant aside from a spike in the years 1979 to 1983. This should not affect our main results

although our estimate ofT F P

may be biased downward during the early 1980s. The data cannotbe extended beyond 2001 since the Exploration and Production unit was consolidated with otherunits in 2002, making the data non-comparable.

11

7/31/2019 MPRA Paper 33024

13/40

Figure 1: Total Factor Productivity of Petrobras and Brazil, 1976-2001

1975 1980 1985 1990 1995 2000 200560

80

100

120

140

160

180

200

Years

1994

=1

00

Petrobras

Brazil

is an indicator that factor shares are similar in the two countries. We use a capital

share of 0.45 and material share of 0.35 11.

Using equation 1 and the Brazilian data we computed Petrobras TFP shown in

Figure 1. (Figure 1 also shows aggregate Brazilian TFP, which will be discussed

below.) During the time that Petrobras was a legal monopolist, there is very little

sustained growth in TFP. There is a deep and abrupt fall in TFP in the second half

of the 1970s with a recovery in the early 1980s. From 1984 until 1994, TFP shows

no sustained gains. On the other hand, during the reform era from 1995 until 2001

TFP almost doubled, growing 95 percent.

This calculation allows us to decompose the sources of growth of labor produc-

tivity (henceforth productivity). Dividing Equation 1 by Nt, then taking logs yields:

11We use the KLEM data set for Oil and Gas Extraction (Industry Group 4) described in Jor-genson and Stiroh [20]. See the Appendix for details on the data.

12

7/31/2019 MPRA Paper 33024

14/40

log

Yt+sNt+s

log

YtNt

/s = [logAt+s logAt] /s+ (2)

log

Kt+sNt+s

log

KtNt

/s+

log

Mt+sNt+s

log

MtNt

/s

Table 1 reports a growth accounting for the pre- and post-reform periods. The

growth rate of productivity more than tripled after the end of monopoly. The sources

of growth in the two subperiods also changed. In the first subperiod, productivity

grew almost completely due to an increases in K/NandM/Nwhile TFPwas nearly

constant. In the second, TFP became the major source of growth.

Table 1 - Growth Accounting of Petrobrass Labor Productivity (%)

Period change in Y/N due to TFP due to K/N due to M/N

1977-1994 4.7 0.3 2.1 2.2

1995-2001 14.6 9.6 2.7 2.3

This result is similar to the the findings of Bugarin, et al. [8] for the aggregate

Brazilian economy. They find that the government encouraged capital accumulation

to keep the economy growing after the oil shocks of the 1970s, despite the lack of

technological progress. The government subsidized private companies and had SOEs,

like Petrobras, increase investment.

We now examine the assumptions used in the construction of the capital and

materials series. We find that they do not significantly impact our results.

There are a number of difficulties that could lead to the mismeasurement of cap-

ital. First, Brazils history of very high inflation rates complicates the measurement

of capital. Since the data reported on an annual basis, the same nominal expendi-

ture will buy much less real capital if it occurs in December rather than January.Brazilian accounting rules required indexation of financial reports during the high

inflation years, which mitigates this problem.

13

7/31/2019 MPRA Paper 33024

15/40

High inflation also led to rapid depreciation of the Brazilian currency, which

affects the conversion to dollars. Using the December exchange rate will tend to

underestimate the real capital stock when inflation is high. The Brazilian currency

was pegged to the dollar. The peg was periodically adjusted (daily during hyper-inflation), depreciating the Brazilian currency. Therefore, the currency would be at

its most depreciated level at the end of the year. Capital expenditures are converted

using the exchange rate when the Brazilian currency buys the fewest dollars during

the year, which will tend to understate investment. In turn, TFP will tend to be

biased upward. This systematic bias is not present in low inflation years. This factor

will lead us to understate the impact of the reform on TFP growth. Since the oil

sector reform occurred after high inflation ended in Brazil (the Real Plan that ending

high inflation was implemented in July 1994), the upward bias will only be present

before the reform.

We use the U.S. oil equipment price index to deflate investment, which raises

the concern that it may not reflect price changes in Brazil. Overall capital price

movements are similar in the two countries. The aggregate price of equipment relative

to the GDP deflator have similar declines in the United States and Brazil from 1976

to 2001. However, U.S. oil equipment prices do not have the same behavior as U.S.

aggregate capital prices, so the aggregate price index may not be appropriate. We

use the U.S. oil equipment index because it is specific to the oil industry and there

is no corresponding Brazilian index. In any case, the results are not sensitive to

the selection of deflator. The results are not significantly altered if the Brazilian

aggregate equipment price index is used as a deflator.

Another issue is that the relative price of structures in Brazil nearly doubled in the

mid-1980s (Bugarin, et al. [7]). This mismeasurement would also tend to overstate

TFP growth since real investment in structures would be understated after the mid-

1980s. However, the quantitative impact is likely to be tiny. Structures make up

very little of Petrobrass capital, less than 2 percent of the value of net property,

plant and equipment in 2005.

Turning to the materials series, the constant materials use per producing wellappears to be a reasonable assumption. Examining U.S. data, we compare TFP

using real materials (from Jorgenson and Stiroh [20]) with TFP using the producing

14

7/31/2019 MPRA Paper 33024

16/40

Figure 2: Alternative U.S. TFP Measures, 1976-2001

1975 1980 1985 1990 1995 2000 200540

50

60

70

80

90

100

110

Years

1994=

100

TFP (Wells)

TFP (Materials)

wells proxy. As can be seen from Figure 2, the two estimates are very similar after

1984. The divergence prior to the mid-1980s can be explained by an increase in well

development during the oil crisis years. This industry does not just produce oil and

gas products, but is also engaged in exploration, drilling and equipping oil and gas

wells. A substantial share of materials use is for well development12. U.S. materials

per producing well are relatively stable except for a spike from 1978 to 1984, which

coincides with a spike of the same magnitude in well development.

Brazilian material use per producing well may have fallen during the 1990s. As

discussed above, material usage is largely driven by well attributes and Petrobras

shifted production to newer and larger wells that tend to use fewer material inputs.

If this is the case, our measure of TFP is biased downward since we overestimate

12The largest categories of expenditures in the U.S. Census of Mineral Industries are Purchasedmachinery installed and Steel mill shapes and forms.

15

7/31/2019 MPRA Paper 33024

17/40

material input.

Figure 3: Petrobras TFP with Materials Proportional to Output and Wells, 1976-2001

1975 1980 1985 1990 1995 2000 200560

80

100

120

140

160

180

200

Years

199

4

=1

00

Proportional to Output

Proportional to Wells

On the other hand, if material used per well increased, we overestimate TFP

growth. However, it is unlikely that our results would be overturned. Material usage

per well would have to have increased by a factor of nearly 7 to account for the

post-reform increase in productivity. TFP (A in Equation 1) grew by a factor of

1.95. Given an input share of 0.35, material usage per well would need to increase

by a factor of 6.74 (1.95 = (6.74)0.35). Petrobras engineers we spoke with did not

believe that such an enormous increase in materials use occurred.

As an additional robustness check, we calculated TFP assuming that growth

rate of material use was the same as that of output

13

. This assumption implies that13We set Mt = Yt, for some > 0, which gives At =

Y(1)t

KtN

(1)t

. Note that cancels out

when T F P is normalized to an index, so the calculation is not affected by the value of .

16

7/31/2019 MPRA Paper 33024

18/40

material use per well increased after the reform since output per well increased. As

can be seen from Figure 3, the path of TFP is largely unchanged and continues to

show an enormous increase after the reform.

5 Sources of Productivity Growth

Since TFP growth increased so dramatically, it is natural to ask what the sources of

that growth were. In this section, we argue that TFP accelerated due to the reform

that brought the threat of competition and privatization after 1995. Though the

increase in TFP coincides with the policy change (Figure 1), it could have resulted

from some other unrelated sources. We first show that the increase did not result from

a number of plausible candidates: improvements in the aggregate Brazilian economy,

technological progress in oil extraction and maturation of previous investments. We

then show that inputs were used inefficiently prior to the policy change. With the

loss of its monopoly, Petrobras quickly reduced its use of inputs while continuing to

expand output. (Output grew at similar rates before and after the policy change).

The company reorganized its structure in an effort to appeal to outside investors.

The evidence is consistent with non-economic goals becoming less important relative

to the economic goal of improving efficiency.

5.1 Ruling Out Potential Candidates

One potential source of the improvement in productivity is the expansion of the

Brazilian economy in the 1990s. It began recovering from its depression of the 1980s

when many reforms such trade liberalization, currency stabilization, deregulation,

and privatization took place. After these reforms, Brazilian labor productivity and

TFP began growing (Bugarin, et al. [9]). This raises the possibility that it was the

aggregate environment that caused the increase in productivity at Petrobras.

The evidence is not consistent with this explanation. We computed the TFP

of Brazil using the methodology used in Bugarin et al [9]. Returning to Figure 1,

TFP for Brazil and Petrobras have very different patterns14. Brazilian TFP started

14Brazils T F P is calculated using a value added measure while Petrobrass uses a gross output

17

7/31/2019 MPRA Paper 33024

19/40

increasing earlier than Petrobrass. While Petrobrass TFP shows sustained growth

after 1995, Brazils begins to decline after 1998. We conclude that Petrobrass expe-

rience is not due to the movements of the aggregate Brazilian economy.

It is also possible the increase was the result of significant improvements in oilextraction technology after 1995. The evidence does not support this explanation.

We compare the Brazilian industry to those of the United States, United Kingdom

and Norway. If technology advanced faster after 1995 the relative productivity should

not change. We would expect the oil industries in other countries to implement the

new technology and experience similar growth in the labor productivity.

Most of Brazils oil comes from offshore wells, accounting for over 85 percent of

production in 2004. Norways and UKs oil industries resemble Brazils since they

extract oil (exclusively) from deep water wells and began offshore production the

mid-1970s. The United States also extracts some oil from deep water but much

comes from land-based sources. Therefore, the comparison should control for both

general technological change in extraction that would affect all countries and specific

change that would only affect offshore extraction.

We do not have enough data to compute TFP for the Norwegian and UK oil

industries or the U.S. industry beyond 1996, so we examine labor productivity.

Figure 4 plots labor productivity, as measured by oil production per worker (in-

dices set to 100 in 1994), for the four countries. While there are clearly differences

in the year to year movements of productivity, the overall pattern is similar prior

to 1995. Productivity falls in the late 1970s and begins to grow in the 1980s. The

magnitude to the growth from the 1980s to 1994 is similar for all four countries.

After 1995, Brazil begins to strongly outperform the other three.

Another possible candidate to explain Petrobrass rapid productivity growth since

1995 could be the maturation of previous investments. There is a lag between explo-

ration and production. In our case, Petrobras could have increased its investment in

exploration for new and more productive fields during the 1980s and the increase in

productivity could be due to the start of production in these fields.

Figure 5 plots the number of exploratory wells drilled and real crude oil pricesmeasure. Data constraints prevent us from calculating a value added measure of Petrobrass T F P.We use a capital share of 0.37 for Brazil.

18

7/31/2019 MPRA Paper 33024

20/40

Figure 4: Oil Extraction Labor Productivity - Brazil, Norway, UK and the UnitedStates (Index 1994=100), 1976-2001

1975 1980 1985 1990 1995 2000 20050

50

100

150

200

250

300

Years

1994=

100

Norway

Brazil

United States

United Kingdom

from 1970 until 2001. Until 1999, the number for the whole industry corresponds to

those drilled by Petrobras. From the late 1980s through the 1990s, there relativelylittle exploration activity. The number of wells drilled is strongly correlated with the

oil prices, which were low during this period. Therefore, there is little evidence of

the increase in TFP being caused by previous exploration coming online.

While the 1980s were not a period of heavy exploration, one might be concerned

that fields that were coming online were particularly high quality. The Campos Basin

contains a number of high quality fields and production has shifted to this area. It

does not appear that the quality of these fields is responsible for the increase in

productivity after 1994. The dominance of Campos precedes the increase in produc-

tivity by a decade. By 1985, Campos accounted for over 60 percent of Petrobrass

oil production and has been slowly increasing its share since then.

19

7/31/2019 MPRA Paper 33024

21/40

Figure 5: Brazil Wells Exploration and Real Crude Oil Price, 1976-2001

1965 1970 1975 1980 1985 1990 1995 2000 20050

20

40

60

80

100

120

140

Year

Num

berofExploratoryWellsDrilled

Number of Wells

2000USDollars

10

15

20

25

30

35

40

45

50

55

60Oil Prices

Finally, the increase in productivity might have been due to one particularly good

field within Campos. We found one potential candidate for such a field in Marlim, a

large field discovered in 1985 that began commercial production in 1994. To measure

the impact of this field, we compared two measures of labor productivity. The first

uses Petrobrass total oil production as the measure of output. Output in the second

is total oil production excluding Marlims production. Both use the same labor

measure, the total employment series used in the TFP calculation since we do not

know how many workers were employed in Marlim. Labor productivity still nearly

doubles, growing 76 percent, between 1995 and 2001 when Marlims production is

excluded. (It grows 174 percent with Marlim.) All the difference comes at the end of

the period when Marlims production reaches its peak. The two measures grow by

nearly the same rate until after 1998. In fact, productivity growth is higher in 1998

without Marlim (54 percent) than with (49 percent). Therefore, the rapid increase in

20

7/31/2019 MPRA Paper 33024

22/40

productivity right after 1994 can not be accounted for by the beginning of Marlims

production. Oil production and productivity grow even when Marlim is excluded

while employment falls sharply. This calculation is a very conservative lower bound,

since we are including labor input used in Marlim but excluding its output. Marlimis a very large field that Petrobras has expended substantial resources developing. At

minimum, labor productivity nearly doubled in less than a decade, so we conclude

that this field is not the reason for increasing productivity.

5.2 Our Explanation: The Reform of 1995

We now turn to our explanation of the increase in productivity. We argue that the

threat of competition and privatization led to a change in managements priorities.

We show that inputs were used more efficiently with the reform. Overstaffing wasreduced and production was shifted to more productive wells. The inefficient use

of inputs likely reflected non-economic goals such as maximizing employment that

became less important after the reform.

We begin by showing that the increase in Petrobrass productivity growth rate

during the reform era is statistically significant. We run a difference in differences

regressions using Norway, the United Kingdom and the United States as controls.

We estimate a pooled panel regression on labor productivity growth of the four

countries. The treatment Reform is a dummy variable that takes the value 1 for

Brazil in the years 1995 to 2001. We correct for country specific autocorrelation andheteroskedastic errors by using Newey-West standard errors15. The sample period

is 1977 to 2001 (the UK data begins in 1979). The dependent variable, the annual

growth rate of labor productivity, is given by the log difference of labor productivity.

The results are shown in Table 1. Column (1) reports our baseline, the regression

with time dummies. The coefficient of the reform is positive and significant at 1

percent level, a strong result especially considering the relatively short treatment

period.

15We implement these regressions using the newey2 command in Stata 9. We set the maximumautocorrelation to 3 years, following the rule of thumb of using the cube root of the number of timeperiods T: T

13 = 25

13 3.

21

7/31/2019 MPRA Paper 33024

23/40

Table 1 - Labor Productivity Panel Regressions

Dependent Variable: growth rate of labor productivity

Equation (1) (2) (3)Reform 0.1137 0.0843 0.0879

(0.000) (0.047) (0.008)

Controls Time, Time, Time,

1995 Treat 1995 Treat 1994 Treat

Country

Obs. 98 98 98

p-value

We did a number of robustness checks. In equation (2), we repeat the estima-

tion of equation (1) but we add country controls. The reform indicator is reduced

somewhat, but is still significant at the 5 percent level.

The results are not sensitive to our dating of the reforms beginning. Since the

event we mark the beginning of the reform (Cardosos election) occurred in late

1994, we set the beginning of the reform in annual data as 1995. However, a small

part of 1994 comes after the reform. In column (3), we add 1994 to the treatment

period. (Reform takes the value 1 for Brazil in the years 1994 to 2001.) The

reform indicator remains positive and significant at the 1 percent level. The reformis associated with an increase in growth labor productivity around 8 percent a year.

Again, we obtain this strong finding despite having a relatively short treatment

period.

As a check that the three countries are reasonable controls, we estimate an other

pooled regression (not reported here) for the years prior to the reform (1978-1994)

with a dummy for Brazil. The Brazil dummy is not significant, which suggests that

Brazils oil extraction labor productivity growth process was similar to the other

countries prior to the reform.

Given the small sample size, we were concerned that we could not be rely on

inference based on asymptotic results. We tested the residuals for normality using

22

7/31/2019 MPRA Paper 33024

24/40

the Skewness-Kurtosis test. It does not reject the null of normality at the 5 percent

level.

To check that the results are not driven by changes in the overall economy, we

rerun the regressions after removing average productivity growth in each country.Since Brazil has a history of macroeconomic instability and significant changes in

economic policy, the results could be driven aggregate changes in Brazils economy.

Table 2 reports the results using the growth rate of the labor productivity in the

oil sector minus the growth rate of GDP per capita (from the Penn World Tables)

as the dependent variable. This approach has the advantage of controlling for the

country specific effects without losing degrees of freedom to country dummies.

Table 2 - Labor Productivity Panel Regressions

Dependent Variable: Difference of growth rate of labor productivity in the oil sector

and aggregate economy

Equation (1) (2)

Reform 0.1447 0.1160

(0.000) (0.001)

Controls Time, Time,

1995 Treat 1994 Treat

Obs. 98 98

p-value

The results are robust to the new measure of productivity growth. Column

(1) reports the results of the baseline specification. The reform indicator remains

positive and significant at the 1 percent level. The reform coefficient is robust to

changes in the year which the reform starts. As reported in column (2), adding 1994

to the reform period generates similar results. These results indicate that the reform

increased the annual growth rate of labor productivity 10% above the growth rate

of the labor productivity of the aggregate Brazilian economy.

23

7/31/2019 MPRA Paper 33024

25/40

How was Petrobras able to increase its productivity so significantly? After the

reform, the company made a number of changes in an effort to become more inter-

nationally competitive. Company managers explicitly state that prior to the reform,

Petrobras pursued a number of non-economic goals such as encouraging Brazilianeconomic development by building infrastructure in remote areas and purchasing

domestically produced inputs (2003 Annual Report). During the 1970s, the com-

pany restricted its foreign purchases to conserve the governments foreign currency

reserves (1977 Annual Report).

Anticipating a loss of market share in Brazil with the loss of monopoly, it changed

its corporate and financial structure so the the company would be able to expand

overseas. A major goal was to be able to raise capital at a cost that was at least

compatible with that of its competitors (2003 Annual Report). It issued a minority

of its voting stock as American Depository Receipts on the New York Stock Ex-

change in 2000, which requires the company maintain U.S. accounting, disclosure

and corporate governance standards.16

The data are consistent with this change in priorities. Employment fell rapidly

after the reform, suggesting there was overstaffing before. Figure 6 shows the num-

ber of employees. Employment began declining prior to the loss of monopoly status,

though its rate of decline increased sharply after 1994. This decline was due to

the reforms implemented by the Collor government that took office in 1990, part of

which was an across the board reduction in public employment. These changes re-

duced the number of employees at Petrobras. However, despite this earlier reduction

of employment productivity does not start increasing until after 1994. This suggests

that the earlier reduction of employment was not enough to change Petrobras man-

agers behavior. Only once the end of monopoly was approved by Congress signalling

that competition and privatization were a real threat did Petrobras implement other

changes that increased productivity.

The data are consistent with Petrobras employing more workers than required to

produce its output, either for specific patronage reasons or to increase employment

16

This offering was not part of a privatization program. In this offering, the Federal governmentreduced its holdings from 82 to 56 percent of voting shares. Recall that it is legally obligated tohold 51 percent of voting shares.

24

7/31/2019 MPRA Paper 33024

26/40

Figure 6: Number of Workers (in 10,000s), 1976-2001

1975 1980 1985 1990 1995 2000 20051

1.5

2

2.5

Years

NumberofWorkers(in10000s)

generally. A number of authors have argued that political considerations led to over-

staffing. Some overstaffing was due to patronage. Geddes [18] argues that Petrobras

began to be used for political purposes in the 1960s, while it had been relatively un-

politicized before. Management power was given to the oil workers union to garner

political support of the employees. Randall [34] estimates that that the company was

overstaffed by 20 percent. Up to 10 percent of the workforce were political patrons,

hired at the behest of the government. Political pressure also forced the rehiring of

fired employees, so both the replacement and original employees were on the pay-

roll. Petrobras was also expected to assist in economic development which may have

encouraged excess employment in an attempt to build up the Brazilian economy.

Oil wells were also used inefficiently, with low quality wells kept in production

prior to the policy change. Figure 7 shows the number of wells in production. Prior

to 1995, the number of wells increased steadily. In 1995, wells in production decline

25

7/31/2019 MPRA Paper 33024

27/40

(sharply) for the first time in the period covered.

Figure 7: Number of Wells in Production, 1975-2001

1975 1980 1985 1990 1995 2000 20051000

2000

3000

4000

5000

6000

7000

8000

9000

Years

NumberofWells

The decline in wells coincides with a large increase in well productivity. As canbe seen in Figure 8, output per well declines in the period prior to the reform. The

development of the productive Campos field led to an increase in the early 1980s.

However, Petrobras added so many wells in less productive areas that total output

per well fell below its pre-Campos level. After the reform, Petrobras seems to have

removed poor wells from production and concentrated its efforts on the best wells.

Marginal wells may have been developed to spread production and the associated

employment, tax revenue and other advantages of local production over a wider

geographical area.

The geographical distribution of wells does change after the reform, with thenumber of wells in less productive areas declining. The number of wells in the States

of Alagoas, Sergipe, and especially Bahia, states with some of the least productive

26

7/31/2019 MPRA Paper 33024

28/40

Figure 8: Output Per Well in Production

1975 1980 1985 1990 1995 2000 200590

100

110

120

130

140

150

160

Years

Thous

andsofBarrelsperYearperWell

wells, fell sharply while production has largely shifted to more productive areas. In

1994, wells in Bahia produced an average of 9,300 barrels a year while wells in theState of Rio de Janeiro produced an average of 402,000 barrels a year. The State

of Rio de Janeiro is where the high quality Campos reserves are located and is a

large source of increased production since the reform. Figure 9 shows the number of

wells in Brazil and the state of Bahia. Prior to the reform, the number of wells in

both were growing at the same rate. After the reform, the number of wells declined

with much of the decline coming in Bahia: In 1995, total wells declined by 509 while

Bahias wells declined by 419. Bahia is a politically important state so maintaining

production there may have been politically motivated.

The rapid change in the use of inputs in the absence of major technical change

suggests that the goals of Petrobrass managers changed with the loss of monopoly

27

7/31/2019 MPRA Paper 33024

29/40

Figure 9: Number of Wells in Production, Brazil and Bahia 1991-2004

1990 1992 1994 1996 1998 2000 2002 200460

70

80

90

100

110

120

Years

1994=

100

Brazil

Bahia

rights. The evidence is consistent with non-economic goals, such as spreading tax

revenue across a wider base, becoming less important relative to the economic goals

of reducing costs and increasing productivity.

6 Competition and Productivity

The reform brought about a remarkable increase in productivity. What makes it all

the more remarkable was how little competition Petrobras faced over the period we

study. Even though Petrobras lost its dejure monopoly, defacto Petrobras is still

a monopolist17 (Palacios [30] and Lewis [25]). The results show that the threat of

17

There are still barriers to entry and the sector is quite risky due to government intervention.For example, recently the Brazilian government blocked a price increase in the domestic market inresponse to the high prices of oil in the international market.

28

7/31/2019 MPRA Paper 33024

30/40

competition, even absent actual competition, can increase productivity. We discuss

the implications of this finding for studying competition and designing reforms.

Petrobras potentially faces two sources of competition in oil extraction: domestic

production and imports. There was little competition from either source.Recall that the reform was slowly phased in. The Petroleum Law was not passed

until two years after Amendment 9 was adopted. The Petroleum Law laid out a

process of gradual opening. It took nearly a decade from the proposal to actual

market opening.

There was very little entry into the oil extraction market. Table 3 shows the

number of fields that Petrobras and other companies bought in each round that took

place since 1998. The last column shows that share of all fields purchased over the

period. Petrobras has at least an interest in nearly three quarters of new concessions.

Even though the number of areas bought by other companies has increased, they have

had little success discovering oil. As can be seen in Table 4, it was not until 2003

that a company aside from Petrobras (Shell) extracted oil in Brazil. In the last line

of Table 4 we show the Herfindhal-Hirschman Index (HHI) (HHI is the sum of the

square of the market share of all companies in a given industry). It takes values

between zero and one, with higher numbers indicating more concentration in the

industry). The HHI has not changed much since the end of monopoly. According

to Kingstone [23], this has been used as evidence that Petrobras kept all the most

promising areas in Round Zero.

Table 3 - Purchase of Rights of Exploitation 1998-2004

29

7/31/2019 MPRA Paper 33024

31/40

Measuring Entry Number of Fields

1998 1999 2000 2001 2002 2003 2004 Share 98-04 (%)

Petrobras 96 1 0 7 3 85 57 54.5

Petrobras with others 0 6 11 7 5 0 50 17.3

Others 19 5 10 19 13 16 47 28.2

Total 115 12 21 33 21 101 154 100

Source: ANP

There was also very little import penetration. The market was not liberalized

until 2001. Petrobras has enormous advantages as the incumbent that cushion it

from foreign competition. It owns all oil terminal facilities in Brazilian ports, though

competing facilities are currently being built.

Table 4 - Petrobrass Share of Oil Extraction 2001-2004 (%)

Period 2001 2002 2003 2004

Extraction 100.0 100.0 99.2 96.6

HHI 1.00 1.00 0.98 0.93

Source: Anuario Estatistico (ANP) and Petrobras.

Petrobras dominates most aspects of the Brazilian oil market. (Lewis [25] and

Ellsworth and Gibbs [14]). It controls nearly all the refining market. Table 5 shows

the percentage of Petrobras in the total refining capacity between 1997 and 2003.

Petrobras has 98% of the Brazilian installed capacity to refine oil during the entire

period. The end of the monopoly did not affect Petrobrass share in the industry18.

Maintaining control of exploration and refining gives Petrobras control of the gasoline

market. There is little international trade in gasoline since it must be formulated18Since 1954, four private companies have operated in the refining sector. They were allowed to

operate since they were operating prior to the creation of Petrobras (Serour [36])

30

7/31/2019 MPRA Paper 33024

32/40

to local standards, which prevents taking full advantage of economies of scale in

transportation.

Table 5 - Share of the Installed Refining Capacity 1997-2003 (%)

Period 1997 1998 1999 2000 2001 2002 2003

Petrobras 98.7 98.7 98.5 98.6 98.6 98.3 98.4

Other Companies 1.3 1.3 1.5 1.4 1.4 1.7 1.6

HHI 0.97 0.97 0.97 0.97 0.97 0.97 0.97

Source: Anuario Estatistico (ANP).

6.1 Policy Implications

The results have a number of policy implications.

First, economists should be cautious when using market shares as a indicator of

competitiveness. Using the HHI, one might conclude that the reform was a failure

since Petrobras did not cede its market share. However, it was successful in increasing

productivity. Studies using market share data will miss threats of competition, whichcan have real effects.

Another implication is that if privatization is not possible, reducing barriers to

competition can increase productivity. There are no plans to privatize Petrobras and

it is unlikely to occur in the future (Kingstone [23]). The removal of the monopoly

guarantee alone generated significant political conflict.

The results lend support to the view that the competitive environment is an

important determinant of productivity, regardless of ownership. In fact, many in-

stances when only ownership was changed have not resulted in an improvement in

performance (Bartel and Harrison [3]). When privatization is not politically viable,increasing competition in the markets of state-owned firms can provide an avenue

for improving performance.

31

7/31/2019 MPRA Paper 33024

33/40

7 Conclusion

We show that competition can be a strong spur to productivity growth. The threat

of competition alone led to a large and swift increase in Petrobrass productivity.The results provide support for the idea that closing off competition in the 1970s

contributed to Brazils poor economic performance in the 1980s. Petrobras is a mi-

crocosm of the aggregate Brazilian economy. As documented in Bugarin, et al. [9, 8],

Brazilian TFP began to fall after the government expanded state-owned enterprises

(including Petrobras) and raised trade barriers during the 1970s as a strategy to keep

the economy growing despite the worldwide recession. While from 1968 to 1974 the

economy grew rapidly due to surging TFP growth, after 1974 TFP began to fall.

Bugarin, et al. [9, 8] argue that Brazils falling TFP in the 1970s was due to the

closing off of competition. The findings give support to this argument. While thestudy of a single industry, even a large one like oil, cannot definitely answer whether

restricting competition reduced TFP, it suggests that this is a fruitful avenue of

inquiry.

References

[1] M. A. Aldeman and C. G. Watkins. Costs of aggregate hydrocarbon reserve

additions. mimeo, MIT, 2003.

[2] Carlos Eduardo Barreto. A Saga do Petroleo Brasileiro: A Farra do Boi. Nobel,

2000.

[3] Ann P. Bartel and Ann E. Harrison. Ownership versus environment: Disentan-

bling the sources of public sector inefficiency. The Review of Economics and

Statistics, 87(1):135147, 2005.

[4] D. B. Bennion, D. W. Bennion, F. B. Thomas, and R. F. Bietz. Injection water

quality - a key factor to successful waterflooding. mimeo, Society of Petroleum

Engineers, 1994.

32

7/31/2019 MPRA Paper 33024

34/40

[5] Timothy F. Bresnahan and Peter C. Reiss. Entry and competition in concen-

trated markets. Journal of Political Economy, 99(5):9771009, 1991.

[6] M. J. Bromley, S. H. Gaffney, and C. E. Jackson. Oilfield emulsion control,techniques and chemicals used to separate oil and water. mimeo, Petrolite

Limited, 1993.

[7] Mirta Bugarin, Roberto Ellery, Victor Gomes, and Arilton Teixeira. Investment

and capital accumulation in Brazil from 1970 to 2000: A neoclassical view.

mimeo, Fundacao Capixaba de Pesquisa, 2003.

[8] Mirta Bugarin, Roberto Ellery, Victor Gomes, and Arilton Teixeira. From a

miracle to a disaster: the Brazilian economy in the last 3 decades. mimeo,

Fundacao Capixaba de Pesquisa, 2004.

[9] Mirta Bugarin, Roberto Ellery, Victor Gomes, and Arilton Teixeira. The Brazil-

ian depression in the 1980s and 1990s. In Timothy Kehoe and Edward Prescott,

editors, Great Depressions of the Twentieth Century. Federal Reserve Bank of

Minneapolis, 2007.

[10] Roberto Campos. A Lanterna na Popa. TopBooks, 1994.

[11] Harold L. Cole, Lee E. Ohanian, Alvaro Riascos, and James A. Schmitz Jr. Latin

America in the rearview mirror. Journal of Monetary Economics, 52(1):69107,January 2005.

[12] C. B. Cummings and C. E. Engelman. The theory and economics of electrostatic

treaters. In SPE Production Operations Symposium. 1989.

[13] Paulo F. de Castro, Victor Gomes, and Marc A. Muendler. Brazils

Establishment-Worker Data Base RAIS (Relacao Anual de Informacoes Soci-

ais). mimeo, UCSD, 2005.

[14] Chris Ellsworth and Eric Gibbs. Brazils natural gas industry: Missed oppor-tunities on the road to liberalizing markets. mimeo, The James A. Baker III

Institute for Public Policy of Rice University, 2004.

33

7/31/2019 MPRA Paper 33024

35/40

[15] Jose E. Galdon-Sanchez and James A. Schmitz Jr. Competitive pressure and

labor productivity: World iron-ore markets in the 1980s. American Economic

Review, 92:12221235, September 2002.

[16] Patricio Garcia, Peter F. Knights, and John Tilton. Measuring labor produc-

tivity in mining. Minerals and Energy, 15(1):3139, March 2000.

[17] Patricio Garcia, Peter F. Knights, and John Tilton. Labor productivity and

comparative advantage in mining: The copper industry in Chile. Resource

Policy, 27(2):97105, June 2001.

[18] Barbara Geddes. Building state autonomy in Brazil, 1930-1964. Comparative

Politics, 22(2):217235, 1990.

[19] Austan Goolsbee and Chad Syverson. How do incumbents respond to the threat

of entry?: Evidence from the major airlines. mimeo, University of Chicago, 2004.

[20] Dale W. Jorgenson and Kevin J. Stiroh. Raising the speed limit: U.S. economic

growth in the information age. Brookings Papers on Economic Activity, 1:125

211, February 2000.

[21] James A. Schmitz Jr. What determines productivity? lessons from the dramatic

recovery of the U.S. and Canadian iron ore industries following their early 1980s

crisis. Journal of Political Economy, 113(2):582625, June 2005.

[22] James A. Schmitz Jr. and Arilton Teixeira. Privatizations impact on private

productivity: The case of Brazilian iron ore. Review of Economic Dynamics,

forthcoming.

[23] Peter Kingstone. The long (and uncertain) march to energy privatization in

Brazil. mimeo, The James A. Baker III Institute for Public Policy of Rice

University, 2004.

[24] John Laitner. Monopoly and long-run capital accumulation. The Bell Journalof Economics, 13(1):143157, 1982.

34

7/31/2019 MPRA Paper 33024

36/40

[25] Steven W. Lewis. Deregulating and privatizing Brazils oil and gas sector.

mimeo, The James A. Baker III Institute for Public Policy of Rice University,

2004.

[26] Gary Libecap and Steven Wiggins. The influence of private contractual failure

on regulation: The case of oil field unitization. Journal of Political Economy,

93(4):690714, 1985.

[27] David McKenzie and Dilip Mookherjee. Distributive impact of privatization

in Latin America: An overview of evidence from four countries. Economia,

3:161218, Spring 2003.

[28] William L. Megginson and Jeffrey M. Netter. From state to market: A survey of

empirical studies on privatization. Journal of Economic Literature, XXXIX:321389, June 2001.

[29] Steve Olley and Ariel Pakes. The dynamics of productivity in the telecommu-

nications equipment industry. Econometrica, 64(6):12631298, 1996.

[30] Luisa Palacios. The petroleum sector in Latin America: Reforming the crown

jewels. Technical Report 88, Les Etudes du CERI, 2002.

[31] Stephan L. Parente and Edward C. Prescott. Monopoly rights: A barrier to

riches. American Economic Review, 89(5):12161233, December 1999.

[32] Charles C. Patton. Pratical considerations in the design of oil field water injec-

tion systems. In NACE Canadian Region Western Conference. 1976.

[33] Maria Clara Prado. A Real Historia do Real. Record, 2005.

[34] Laura Randall. The Political Economy of Brazilian Oil. Praeger, 1993.

[35] Maria Alejandra Rondon. Dynamic simulation of a vertical two phase separator.

Masters thesis, University of Tulsa, 1999.

[36] Tatyana Serour. Impactos da desregulamentacao sobre o desempenho do setor

petrolifero brasileiro. Masters thesis, Faculdades Ibmec, 2003.

35

7/31/2019 MPRA Paper 33024

37/40

[37] Jean Tirole. The Theory of Industrial Organization. MIT Press, 1994.

[38] M. Steve Worley and Lawton L. Laurence. Oil and gas separation is a science.

mimeo, Society of Petroleum Engineers, 1995.

36

7/31/2019 MPRA Paper 33024

38/40

A Data

The main source for Brazilian data is the Oil Report (Relatorio do Petroleo) from

Ministry of Mines and Energy.

Oil production thousands barrels per day:

1. Brazil: Oil Report, several years (1954-1990). Anuario Estatstico Brasileiro

do Petroleo e do Gas Natural, Agencia Nacional do Petroleo, several years

(1990-2003).

2. United States: Energy Information Administration, Annual Energy Re-

view, 2001, p. 129.

3. United Kingdom: BP Statistical Review of World Energy, June 2003.

4. Norway: Statistics Norway, Oil and Gas Activity, 4th Quarter 2002, Ta-

bles 23 and 24.

5. World production, includes crude oil, shale oil, oil sands and NGLs (nat-

ural gas liquids - the liquid content of natural gas where this is recovered

separately): BP Statistical Review of World Energy, June 2002.

Employment oil and gas extraction, and oil and gas extraction services:

1. Brazil: Oil Report and RAIS. Employment has been adjusted for con-

tracting out. Employment by other firms in the oil extraction industry

are added to Petrobrass exploration and production employment. Prior

to the reform, all these firms were contractors for Petrobras. Beginning in

the late 1990s when other firms could begin oil exploration, employment

will include employees of some firms that are not Petrobras contractors.

For details about RAIS see De Castro, Gomes and Muendler [13].

2. United States: Production and Service Jobs and Wells Drilled. U.S. De-

partment of Labor, Bureau of Labor Statistics, National Employment,Hours, and Earnings (www.bls.gov). Oil and gas production (eeu10131001),

and oil and gas services (eeu10138001).

37

7/31/2019 MPRA Paper 33024

39/40

3. United Kingdom: Employees extraction of mineral oil and natural gas:

SIC 92 CA 11. Department of Trade and Industry, UK.19

4. Norway: Statistics Norway, Oil and Gas Activity, 4th Quarter 2002, Table

45.

Oil Prices prices for oil: Dubai, oil spot crude price. U.S. dollars per barrel.

1972-1985: Arabian Light; 1986-2001: Dubai. Sources: Brazilian Oil Report

and BP Statistical Review of World Energy, June 2002.

Natural Gas Prices prices for natural gas: U.S. Natural Gas Wellhead Price

(Dollars per Thousand Cubic Feet). Source: Energy Information Administra-

tion, Historical Natural Gas Annual.

Investment For investment series, in U.S. dollars, we use the number from the

Oil Report (several years). The nominal series is converted to 1994 dollars

by deflated by the producer prices Oil and gas field machinery and equip-

ment manufacturing. U.S. Department of Labor, Bureau of Labor Statistics

(pcu333132333132). This series begins in 1965. For 1954 to 1964, we use the

U.S. Consumer Price Index.

Wages U.S.: Average weekly earnings of production workers, oil and gas extrac-

tion. U.S. Department of Labor, Bureau of Labor Statistics (ceu1021100004).

A.1 Capital Stock

We constructed the capital stock of Petrobras using the perpetual inventory method

on investment (x). The law of motion for capital stock is:

kt+1 = (1 )kt + xt (3)

where k is the capital stock and is a constant depreciation rate. We assume a

depreciation rate () of 4% per year. Since we start accumulating investment in19Thanks to Philip Beckett and Mike Earp, Department of Trade and Industry, UK.

38

7/31/2019 MPRA Paper 33024

40/40

1954, the year Petrobras was established, the initial capital stock comes from the

data.

A.2 Total Output: Oil and Gas

Typically, petroleum exploration and development yields a joint product: oil and

natural gas. The typical way to aggregate oil and gas output has been to convert

natural gas to oil equivalent at a fixed ratio based on physical thermal content or

on some thermal value content implied by relative wellhead prices at a given point in

time. There is a major problem with fixed coefficient. Relative values of oil and gas

change over time and this affects the problem of firm (see Aldeman and Watkins [1]).

To adjust our data to this possibility, we transform the amount of gas produced

into oil using the price of gas relative to oil in the U.S. spot market. Data constraintsprevent us from producing country specific series. Though the relative prices can

differ across countries, the variances are very similar. The intuition behind these

procedure is that in equilibrium the marginal rate of transformation of gas in oil is

equal to the relative price. Therefore the total amount of oil produced is given by:

yt = ot +pgtpotgt (4)

Where ot is the production of oil in period t (thousand barrels day), gt is the produc-

tion of gas in period t (million cubic feet day), pot is the price of oil (Dubai barrelprice), pgt is the price of natural gas (U.S. natural gas wellhead price dollars per

thousand cubic feet).

39